Static, never active
SkillLock reads text patterns only. It never runs shell commands, scripts, downloads, or installers from what you scan.
AGENT SKILL SECURITY
SkillLock scans instructions and scripts for risky permissions, data transfers, and secret access — without running any code.
Scan a Skill freeSTEP 1
Paste a public or connected private GitHub repository, branch, folder, or file link. Folder links scan only that folder; file links scan only that file. Private scans require the SkillLock GitHub App with read-only access to the repository you choose. Up to 100 supported text files are read; no code is executed.
Read the Skill instructions and text files without executing them.
Show what it may read, send, or run.
No “100% safe” promise — just clear evidence for your decision.
BUILT FOR REAL WORKFLOWS
Before an AI agent follows a new Skill, SkillLock gives you a quick, readable security review of the instructions and included text files.
SkillLock reads text patterns only. It never runs shell commands, scripts, downloads, or installers from what you scan.
Each result separates direct behavior, context that needs a review, and low-risk information so you can focus quickly.
Scan folders on your device, public GitHub links, or a private repository you explicitly connect through the read-only GitHub App.
FAQ
No. It finds known risky patterns and explains the evidence. Review important Skills yourself before installing them.
No. A local folder scan happens in your browser. Only GitHub-link scans ask the server to fetch supported text files from that GitHub repository.
Only repositories you choose when installing the SkillLock GitHub App. The app has read-only Contents permission and cannot change your repositories.
SkillLock is designed for agent Skills used with Codex, Claude Code, and Cursor. It can also review supported text files in a GitHub repository.